White House Clears Path for Federal Agencies to Use Anthropic Mythos AI

The White House Office of Management and Budget is establishing security protocols to allow major federal agencies to begin using Anthropic's Claude Mythos Preview, the powerful AI model that has rattled cybersecurity experts and financial regulators on both sides of the Atlantic since its limited release earlier this month. Federal CIO Gregory Barbaccia informed Cabinet departments of the plan in a memo reviewed by Bloomberg.

The move comes even as government officials grapple with the model's potential to sharply increase cybersecurity risk, and despite an ongoing legal and political clash between the Trump administration and Anthropic over the use of the company's AI technology. The decision to proceed reflects the administration's calculation that the defensive value of Mythos, deploying it to audit and harden federal infrastructure, outweighs the risks of broader access.

Federal Deployment | OMB Security Framework for Mythos Access

The OMB protocols will govern how agencies request, access, and operate Mythos Preview within federal IT environments. Barbaccia's memo to Cabinet departments frames the rollout as a controlled expansion under defined security guardrails, rather than an unrestricted deployment. The approach parallels Anthropic's rollout to UK financial institutions and the Project Glasswing initiative that gave JPMorgan Chase and other systemically important banks a structured path to deploy Mythos offensively against their own systems.

The federal green light comes as the ECB pursues its own systemic risk inquiry and the IMF has warned of catastrophic operational risk from AI-boosted cyberattacks. The US government's decision to authorize internal use while international regulators are still conducting risk assessments underscores how sharply views diverge on whether offensive AI capabilities are better managed through access or restriction.